United States
                    Environmental Protection
                    Agency
Risk Reduction
Engineering Laboratory
Cincinnati OH 45268
                    Research and Development
EPA/600/S2-90/050  Feb. 1991
& EPA       Project  Summary
                    Standard  Test Procedures for
                    Evaluating  Leak Detection
                    Methods:  Pipeline  Leak
                    Detection Systems
                    Joseph W. Maresca, Jr., Robert M. Smedfjeld, Richard F. Wise,
                    and James W. Starr
                       This report presents a standard test
                    procedure  for evaluating the perfor-
                    mance of leak detection systems for use
                    in the pipelines associated with under-
                    ground storage tanks. The test proce-
                    dure is designed to evaluate these
                    systems against the performance stan-
                    dards of the U.S. Environmental Protec-
                    tion  Agency's (EPA's) underground
                    storage tank (UST) regulations (40 CFR
                    Part  280, Subpart D), which cover an
                    hourly test, a monthly monitoring test,
                    and an annual line tightness test. The
                    test procedure can be used to evaluate
                    any type of system that is attached to
                    the pipeline and that monitors or mea-
                    sures either flow rate or changes in
                    pressure or product volume. This proce-
                    dure can be used to evaluate leak de-
                    tection systems that can relate the
                    measured output quantity to leak rate (in
                    terms of gallons per hour) and systems
                    that use an automatic preset threshold
                    switch. The test procedure can be used
                    to evaluate systems used to test pres-
                    surized pipelines or  suction pipelines
                    that are pressurized for the test. The test
                    procedure offers five options for collect-
                    ing the data required to calculate perfor-
                    mance. The results of the evaluation are
                    reported In a standard format on forms
                    provided in the appendices of the report
                    summarized here.
                       This Project Summary was devel-
                    oped by EPA's Risk Reduction Engi-
                    neering Laboratory in Cincinnati, OH, to
                    announce key findings of the research
                    project that is fully  documented in a
                    separate report of the same title (see
Project Report ordering Information at
back).

Introduction
   The EPA's regulations for underground
storage  tanks require owners and opera-
tors to check for leaks on a routine basis
using one of a number of detection meth-
ods (40 CFR  Part 280, Subpart  D). To
ensure the  effectiveness of these meth-
ods, the EPA has set minimum  perfor-
mance standards for equipment used to
comply  with the  regulations.  Deciding
whether a system meets  the standards
has not  been easy, and the EPA will not
test, certify,  or approve  specific brands of
commercial  leak detection  equipment. In-
stead, the EPA has developed  and pub-
lished a series of standard test procedures
that describe how equipment should be
tested to prove that it meets the standards.
Each document on a type of system or
method  explains how to conduct the test,
how to perform the required calculations,
and how to report the results. The results
from each standard test procedure provide
the information needed by tank owners
and operators  to determine whether the
method meets the regulatory requirements.
The final report summarized here is part of
the series published by the EPA.
   The performance results are reported
in terms of leak rate (in gallons per hour),
probability of detection (PD), and probabil-
ity of false alarm (PFA).  The protocol ad-
dresses  the performance  of these leak
detection systems for the leak rates, PD, and
PFA specified in the EPA regulation. The
protocol covers all of the internal EPA
                                                                    Printed on Recycled Paper

-------
release detection  options for piping  but
does not cover the external leak detection
options (those for vapor and groundwater
monitoring). Separate protocols have been
developed for these external systems.
Common types of leak detection systems
that can be  evaluated with the protocol
summarized  here include systems  that
measure pressure, volume,  or flow-rate
changes in the pipeline. Both pressurized
and suction piping systems are addressed,
and if  release detection is required for a
suction system, it is assumed that the  line
will be pressurized for the test.
   The EPA regulation requires two types
of leak detection  tests for underground
pressurized  piping  containing petroleum
fuels. First, underground piping  must be
equipped with an automatic line leak  de-
tector that will alert the operator to  the
presence of a leak by restricting or shut-
ting off the flow of the regulated substance
through the piping or by triggering an audi-
tory or visual alarm. The automatic  line
leak detector must be capable of detecting
Isaks of 3 gal/h defined at a line pressure
of 10 psi within an hour of the occurrence
of a leak and with  a P? of 95% and a PFA
of 5%. The test is designed to detect the
presence of very large leaks that may oc-
cur between  regularly scheduled checks
with the more accurate monthly monitoring
tests or annual line tightness tests. Many
of these systems use a preset-threshold
switch.
   Second, the regulation also  requires
either an annual  line tightness test or a
monthly monitoring test.  The annual  line
tightness test must be capable of detecting
a leak  as small as 0.1 gal/h (defined at a
pressure that is 150%  of the operating
pressure of the line) with a PD of 95% and
a PFA of 5%. The monthly monitoring test
evaluated by this protocol must be capable
of detecting leaks as small as 0.2 gal/h
(defined at the operating pressure of  the
line) with a Pp of 95% and a PFA of 5%. This
monthly monitoring test can be satisfied by
the use of any type of pipeline leak detec-
tion system (line pressure monitor, auto-
matic shutdown line leak detector, etc.)
that conducts a precision test on the pipe-
line system and that can satisfy the perfor-
mance requirements.
   The evaluation protocol  requires that
the performance characteristics of the in-
strumentation be estimated  and  that  the
performance in terms of leak rate,  PD, and
PFA be determined for the specified pipe-
line  configuration and  a wide  range of
product temperature conditions. The prob-
ability of false alarm is  estimated at the
threshold used by the manufacturer (the
threshold being the value at which a leak is
declared), and the probability of detection
is estimated at the leak rate specified  in
the EPA regulation. With one slight differ-
ence,  the  same procedure is  used to
evaluate the performance of the monthly
monitoring  test, the annual line tightness
test, and the  hourly automatic line leak
detection test. For the monthly monitoring
test, the probability of detection is estimated
at a leak rate  of approximately 0.2 gal/h,
while for the line tightness test the prob-
ability  of detection is estimated at a leak
rate of approximately 0.1 gal/h; a 3-gal/h
leak is used in the hourly test.

Options for Estimating
Performance
   A complete specification of system per-
formance requires a statement of the PD at
a defined leak rate, a statement of the PFA,
and an estimate of the uncertainty of the
PD and  PFA.  The performance estimate
should be made over the range of condi-
tions under which the system will actually
be used. They can be made from a perfor-
mance model based on the histograms of
the noise and the signal-plus-noise. The
actual calculations will be made with an-
other representation of the histogram called
the cumulative frequency distribution.
   To estimate the performance of a pipe-
line leak detection system, one must de-
velop  histograms of the  noise  and the
signal-plus-noise. Each histogra.m gener-
ated according to this protocol requires a
minimum of 25  independent tests. This
number ensures that an estimate of the PD
of 95% and the  PFA of 5% can be made
directly from the data and that the uncer-
tainty in the estimate of the PD and PFA, as
measured by  the 95% confidence inter-
vals, is approximately 5%.
   This protocol provides five options for
generating the data necessary  to develop
noise  and  signal-plus-noise histograms.
The first option is to conduct the evaluation
at an instrumented test facility s,pecifically
designed to evaluate pipeline leak detec-
tion systems, and the second is to do it at
one or more operational LIST facilities that
are specially instrumented for the evalua-
tion. Both of these options require that the
data be collected under a specific set of
product temperature conditions, which are
measured as part of the test procedure,
and on a pipeline system that hais defined
characteristics.  The instrumentation  is
minimal and does not require that tem-
perature sensors be placed  inside  the
pipeline. The next two options require that
data be collected over a period of 6 to 12
months, either at 5 operational  LIST facili-
ties where the integrity  of  the  pipeline
systems  has  been  verified,  or at  10 or
more operational LIST facilities. The sta-
tions should be geographically located so
as  to  represent different  climatic condi-
tions. Each of the operational UST facilities
selected  should receive a  delivery of
product to the tank at least once per week.
Options 3 and 4 should provide approxi-
mately  the  same  range of temperature
conditions  specified  in Options 1 and 2
because of seasonal variations in the tem-
perature of the ground and the temperature
of the product delivered to the tank. In the
fifth option, a simulation is used to estimate
the  performance of  the  leak  detection
system. This simulation is developed from
experimentally  validated mathematical
models of all the  sources of noise  that
affect the performance of a particular sys-
tem.

Generating the Noise Histogram

    The primary source of noise for a pipe-
line leak detection system is the thermal
expansion and contraction of the product
in the line. Thus, the performance of most
pipeline leak detection systems is  con-
trolled primarily by temperature changes in
the  product  that is  in the line. These
changes are present unless no product
has been pumped through the pipeline for
many hours. To take  these changes into
account, the protocol requires that all leak
detection  systems be evaluated under a
wide range of temperature conditions.
    The range of temperature conditions
used in this protocol  is based on the re-
sults of an analytical study of the climatic
conditions  found throughout the United
States.  The study estimated the average
difference in temperature, AT, between the
product in the tank and the temperature of
the ground around the pipe.  The results
indicated that values of ±25°F would cover
a wide range of conditions. All systems will
be evaluated in accordance with their own
test protocols under the matrix of tempera-
ture conditions given in Table 1. The  pro-
tocol in this document describes specifically
how to create these conditions.
    Table 1 summarizes the number of tests
that must be done for  each of the nominal
conditions for which  histograms must be
generated.  A temperature  condition is
generated by circulating product for 1  h or
longer at one temperature through a pipe-
line system surrounded by backfill and soil
at another temperature. It is assumed that
the temperature conditions within the range
of each 10°F increment will  be as  uni-
formly distributed as possible. This is  par-
ticularly important for the  conditions
centered on 0°F; about half of the condi-
tions  should  be positive and about  half
should be negative.

-------
 Table 1.  Number of Tests Required for Each Range of Temperature Conditions
     Number of       Percentage          Range of Temperature
       Tests           of Tests              Differences AT?F)
1
4
5
5
5
4
1
4
16
20
20
20
16
4
AT<-25
-25£&T<-15
-15<.&T<-5
-5Z&T<+5
+5ZAT<+15
+15£&T<+25
A7S +25
    The pressure  and volume changes
 produced by the  thermal expansion or
 contraction of any trapped vapor in the
 line also affect the performance of most
 detection systems; in some  instances, a
 leak detection device will simply not work
 if vapor is trapped in the line. It is as-
 sumed in this protocol that the leak detec-
 tion system being evaluated would require
 that the line  be tested for vapor and, if
 vapor were found to be present, would
 either cancel the test or require that the
 trapped vapor be removed before the test
 was begun. Because the protocol makes
 this assumption, evaluators should  en-
 sure  that, when the evaluation is  con-
 ducted at an instrumented test facility (i.e.,
 Options 1, 2, and  5), all vapor has been
 removed from  the pipeline  for all tests
 used in estimating performance. To assess
 the sensitivity of the system to trapped
 vapor, this protocol requires a minimum
 of three tests  with  a small volume of
 trapped vapor in the line. The results of
 these three tests will not be included in
 the  performance estimates  but will  be
 presented in the evaluation report so that
 the manufacturer's claims about the effects
 of trapped vapor on  the test results can
 be better assessed.

 Generating the SIgnal-plus-nolse
 Histogram

   A histogram of the signal-plus-noise is
 a requirement for  estimating the PD for
 each leak rate of interest. The threshold
 value is used to determine the PD directly
 from the histogram of the signal-plus-noise
 for  a given leak rate. A separate histo-
 gram of the signal-plus-noise is required
 for each signal  (i.e.,  leak rate) for which
the performance in terms of PD is desired.
 For each leak rate of interest,  the histo-
 gram of the  signal-plus-noise must be
 developed over the  same temperature
conditions  and  pipeline configurations
 used  to generate  the noise histogram.
This protocol requires, at a minimum, that
the PD be estimated against the leak rate
specified in the EPA regulation for the
 type of leak detection system being evalu-
 ated (i.e., 0.1, 0.2, or 3.0 gal/h).
    Generating the signal-plus-noise histo-
 gram may be simple or may involve signifi-
 cant effort. There are two options. The direct
 approach is to develop  the histogram by
 generating a leak in the line and conducting
 a large number of leak detection tests un-
 der the same conditions used  to develop
 the histogram of the noise. This direct ap-
 proach can be used regardless of whether
 the leak detection system uses a preset
 threshold or measures the flow rate directly.
 Noise and signal-plus-noise histograms are
 required for each temperature condition. In
 this approach, the  histogram  of the
 signal-plus-noise is measured  directly for
 the leak rate at which the PD is desired, and
 thus the relationship between  signal and
 noise is determined directly. If the leak de-
 tection test is short, the data necessary to
 develop  the  noise and  signal-plus-noise
 histograms can be acquired by  conducting
 two tests in succession. The direct approach
 is most beneficial when a PD is required for
 only a few leak rates; otherwise, the time
 required to collect the  data can be exces-
 sive. This approach  is easy to implement
 when data are collected at an instrumented
 test facility or one or more instrumented
 operational UST facilities, but it is cumber-
 some if the data must be collected over an
 extended period at many noninstrumented
 operational UST facilities. If there is a large
 number of leak rates, each requiring an
 estimate of the probability of detection, or if
 the test duration is sufficiently long that only
 one leak detection test can be  conducted
 for a given temperature condition, the sec-
 ond approach would be more logical.
   The second approach is to develop a
 signal-plus-noise histogram from the histo-
 gram of the noise by developing a theoreti-
 cal relationship between the signal and the
 noise.  An experimentally validated  model
that gives the relationship between the sig-
 nal and each source of noise must be de-
veloped. With this model and the histogram
of the  noise, the signal-plus-noise histo-
gram can be developed for any leak rate,
and an estimate of the PD can also be made
 for any leak rate. This relationship must be
 valid over the range of test conditions and
 pipeline configurations  covered by the
 evaluation.  It can  be used with all five of
 the options for data collection. It is particu-
 larly useful for evaluating the performance
 of leak detection systems that require long
 tests or long waiting periods or that ac-
 quire the noise  data at many operational
 UST facilities over a long period of time.

 General Features of the
 Evaluation Protocol
    The general  features of the evaluation
 protocol,  including how the pipeline con-
 figuration affects performance and the 13
 steps  required to  conduct the evaluation,
 are summarized below.

 Pipeline Configuration

    There is  a wide range of pressurized
 pipeline systems that must be tested peri-
 odically for leaks,  and leak detection sys-
 tems must comply  with the EPA regulation.
 The performance  of  many pipeline leak
 detectors, especially pressure detection
 systems, will vary according to the configu-
 ration  of the pipeline system. The magni-
 tude of the signal as well as that of the
 noise  will be affected. This occurs  be-
 cause the overall  compressibility charac-
 teristics  of the  pipeline  system are
 influenced by the choice of material (fiber-
 glass or steel), the use of flexible hosing
 (and its length), and the presence of ap-
 purtenances on  the  line. This interaction
 between the pipeline and the performance
 of the leak  detection system presents a
 challenging problem: the same leak detec-
 tion system can  perform very well on one
 pipeline system and  poorly on  another.
 Fortunately, the  compressibility character-
 istics of the line can be described by the
 bulk modulus of the pipeline system. Two
 pipelines may have different configurations
 but may  have the same compressibility
 characteristics. In  this protocol, the  bulk
 modulus, which can be readily measured,
 is used to characterize the pipeline used in
 the evaluation.
    Pipelines constructed at special instru-
 mented test facilities should simulate the
 important  features  of the type of pipeline
 systems found at  operational UST facili-
 ties. This  protocol  assumes that the  leak
 detection systems  to be evaluated are in-
tended for use  on underground storage
tanks that are typically 10,000 gal in ca-
 pacity, where the diameter of the pipe is
typically 2 in. and the length is usually less
than 200 ft. If the leak detection system will
be used on pipelines with larger diameters
or longer lengths, the evaluator should use
a  proportionately  larger  pipeline in

-------
conducting the evaluation. Whether the
evaluation is conducted at a special instru-
mented testing facility or at one or more
instrumented operational UST facilities, the
requirements are as follows.
    The pipeline, which can be constructed
    of either fiberglass or steel, must have
    a diameter of at least 2 in. ± 0.5 in.
    The pipeline  must be at least 75 ft
    tang.
•   The pipeline  system  should have a
    bulk modulus of approximately 25,000
    psi± 10,000 psi.
•   A mechanical line leak detector must
    be  present within the line if the leak
    detection system being  evaluated
    normally conducts a test with this de-
    vice in place.
    There must be a way to pressurize the
    pipeline system.
    There must be a tank or storage con-
    tainer to hold product withdrawn from
    the line during a test.
    There  must be  a pump to  circulate
    product from the storage  container
    through the pipeline for up to 1 h. (At
    operational UST facilities and at most
    test facilities, this container will be an
    underground storage tank, and a sub-
    mersible pump will be used to pres-
    surize  the  pipeline and   circulate
    product through it.)
•   The pipeline must  have valves that
    can be  used to isolate  it  from the
    storage tank and the dispenser. These
    valves must be checked for  tightness
    under the maximum  operating pres-
    sure of the pipeline system.
•   The pipeline must contain  a petro-
    leum product,  preferably  gasoline,
    during the evaluation.
•   In addition, when an evaluation is done
    at a special test facility, there must be
    a unit to heat or cool the product in the
    storage container.
   When the evaluation is done  at five or
more operational UST facilities that are
geographically separated, it will  suffice if
only one of the facilities meets these crite-
ria, with the exception of the bulk modulus
criterion, which does not have to be met by
any of the facilities.
   The performance of some of the sys-
tems that can be evaluated with this proto-
col will decrease as the diameter  and/or
length of the pipeline increases. This is
particularly true  for  volumetric measure-
ment systems that are directly affected by
thermal expansion  or contraction  of the
product in the pipeline. The performance
estimate generated by this protocol is con-
sidered valid if the volume of the product in
the pipeline system  being tested  is less
than twice the volume of product in the
pipeline used in the evaluation. This is an
arbitrary limitation because it does not take
into account the type of system, the method
of temperature compensation,  or the ac-
tual  performance of the system. It was
selected to allow flexibility in the applica-
tion of the system. Thus, in selecting the
length of the pipeline to be used  in the
evaluation one should consider how the
system will ultimately be used operation-
ally. Because the limitation is arbitrary, this
protocol also allows the manufacturer to
present a separate written justification indi-
cating why the method should  be consid-
ered applicable to pipelines having twice
the capacity, or  more, of the one used in
the evaluation. Concurrence with this jus-
tification must be given by the evaluator.
Both the written justification and evaluator's
concurrence must be  attached to the
evaluation report.

Conducting the Evaluation

   A 13-step procedure is used to conduct
an evaluation. The particulars of the evalu-
ation procedure  depend on
    which performance standard the sys-
    tem will be evaluated against (i.e.,
    hourly test at 3 gal/h, monthly monitor-
    ing test at 0.2 gal/h, or line tightness
    test at 0.1 gal/h)
    whether the leak detection system
    measures the flow rate and uses it to
    determine whether the  pipeline  is
    leaking or uses an automatic  preset
    threshold switch and does not directly
    measure and report flow rate.
   The protocol can be used to evaluate
systems that require multiple tests as well
as those based  on a single tost.
   Step 1—Describe the leak detection
system. Specifying the important features
of the leak detection system is important
for three reasons. First,  a brief description
will identify the system as the one that was
evaluated. Second, changes to the system
may  be made  at a  later date, but the
manufacturer may not feel that the changes
are important enough for him to rename
the system. Such changes may affect the
performance, either for better or worse. If
the characteristics of the system have been
specified in a brief descriptive statement,
the  owner/operator of an underground
storage tank system  will  have a way to
determine whether the detection system
he is  using is actually  the one that was
evaluated.  Third, the owner/operator will
be able to interpret the results of the evalu-
ation more easily if he has this information.
   The description of the leak detection
system need not be excessively detailed,
and proprietary information about the sys-
tem is not required. The description should,
however, include the important features of
the instrumentation, the test protocol, and
detection criterion. If the system requires
multiple tests  before a leak is  declared,
this should be clearly stated. (A summary
sheet on which to describe the  system is
provided in the final report.)
   Step 2—Select an evaluation option.
It must be determined which one of the five
evaluation options will be used: test facil-
ity, one or more instrumented operational
UST facilities,  6- to  12-month data collec-
tion effort at 5  operational UST facilities at
which pipeline integrity has been verified,
6- to 12-month data collection effort at 10
or more operational UST facilities, or vali-
dated computer simulation.
   Step 3—Select temperature and leak
conditions for the evaluation. The tem-
perature and leak conditions must be de-
termined. If the evaluation is done at a test
facility,  at one or more instrumented op-
erational UST facilities,  or by  computer
simulation, the temperature conditions
necessary to compile the noise  histogram
will be developed according to a test ma-
trix, which is  generated before the data
collection begins, and  will be verified by
means of specific diagnostic ground and
product measurements made immediately
before the test. A matrix of leak conditions
will also be generated so that a  histogram
of the signal-plus-noise can be  compiled;
the type of test matrix will  depend on
whether the leak rates are known a priori
or whether a  blind-testing procedure is
used. The protocol is designed to minimize
any advantages that the test crew  might
have because of its familiarity with the test
conditions described in the protocol. Thus,
the performance estimates should be
identical regardless of whether the test
conditions were known a priori.  Two blind
testing techniques are provided; these can
be implemented most easily at an  instru-
mented test facility.
    If the data are collected at operational
UST facilities over a  period of 6 to 12
months, temperature  conditions do  not
need to be artificially generated, but the
relationship between the measured quan-
tity and the flow rate that would be pro-
duced by  a leak at the manufacturer's
standard test  pressure (i.e., the relation-
ship between the signal and the  noise)
should  be defined  and  provided by  the
manufacturer  before the system is evalu-
ated. This relationship is used to generate
the signal-plus-noise  histogram from  the
noise histogram at the EPA-specified leak
rate.  The  relationship can be either a
theoretical one that has been  validated
experimentally or an empirical one that
has been developed through experimenta-
tion.


-------
   Step 4—Assemble equipment  and
diagnostic instrumentation. The proto-
col specifies  certain equipment, appara-
tuses, and measurement systems to be
used in the evaluation. Although these must
be assembled and calibrated, none is par-
ticularly complex or  sophisticated. A de-
scription of each is provided. The protocol
allows for the use of other equipment not
specified by this protocol provided it has
the same functionality and performance as
the equipment described.
   Step 5—Verify the integrity of the
pipeline system. Conducting  a perfor-
mance evaluation of  a leak detection sys-
tem requires  a nonleaking pipeline. If the
pipeline is not tight, the performance of the
system being evaluated will be degraded.
For all but one of the evaluation options
presented in this protocol (Option 4), it is
recommended, though  not required, that
the integrity of the pipeline be verified be-
forehand by  means  of a leak detection
system whose performance is already
known.
   Step 6—Determine the  characteris-
tics of the pipeline system. It must be
determined whether  the  pipeline system
used in the evaluation meets  the minimum
specified conditions. The same pipeline
configuration  can be used regardless of
whether the evaluation is done at a test
facility, one or more  instrumented opera-
tional LIST facilities,  or by the simulation
approach. The compressibility of the pipe-
line system must be within a specified
range; if it is not, a mechanical device can
be used to modify the compressibility char-
acteristics of  the line for  the test. An ex-
ample of  a device that can be used to
modify the compressibility characteristics
of the pipeline system is described in the
final report.
   Step 7—Evaluate the performance
characteristics of the sensor sub-
systems. The resolution,  precision, accu-
racy, and minimum detectable quantity of
the measurement subsystems (instrumen-
tation), as well as what the instrumentation
is measuring  (i.e., specificity),  should be
determined. Also, the  flow  rate at the
threshold should be determined. Although
this step is not actually required in order for
an evaluator to estimate the  performance
of the system, it serves two important pur-
poses. First, it indicates, before the evalu-
ation   is  performed,  whether  the
instrumentation is working according to the
manufacturer's specifications. If the instru-
mentation is not performing properly or if it
is out of calibration, the evaluation should
not proceed until the problems are rem-
edied.  Second, the instrumentation will ul-
timately limit the performance of the leak
detection  system.  If it is evident that the
performance expectations of the manufac-
turer are  more than the instruments will
allow, the evaluation can be stopped be-
fore too much time has been invested or
too much expense incurred. Furthermore,
this step can be completed quickly.
   Step 8—Develop (if necessary) a re-
lationship between the leak and the
output of the measurement system, if the
relationship between the leak and the out-
put of the measurement system (i.e., be-
tween the signal and the noise) is known
or has been supplied by the manufacturer
and  no  direct estimate of the  signal-
plus-noise histogram at the  EPA-specified
leak rate  has been made as part of this
protocol, experiments must be conducted
to verify the relationship. This step is not
necessary if the test  matrix requires  25
tests at the EPA-specified leak rate (i.e.,
developing the signal-plus-noise histogram
with the direct approach).
   Step 9—Develop a histogram of the
noise. A histogram of the noise under the
temperature conditions specified in Step 3
for the pipeline system specified in Step 6
must be developed. This histogram, which
is needed to  estimate the probability of
false alarm, is generated from one or more
pipeline tests, conducted according to the
manufacturer's protocol, for each condi-
tion given in Step 3. If the system uses a
multiple-test procedure, two histograms are
required. The performance of the system,
which includes the entire multiple-test se-
quence, is generated from the data ob-
tained from the  test  that is used  to
determine whether the pipeline is leaking
(in many instances these are the data from
the last test in the sequence). Step 9 is the
heart of any evaluation. Once the histo-
gram of the noise is known  and either the
relationship  between the signal and the
noise is  known or a  histogram  of the
signal-plus-noise has been developed, the
performance  of the system can be esti-
mated.
   Step 10—Develop a histogram of the
signal-plus-noise. A histogram  of the
signal-plus-noise for each leak rate at which
the system will be evaluated and under the
same conditions used to generate the noise
histogram must be developed. If system
uses a multiple-test procedure, two histo-
grams are required. The performance of
the system,  which includes the entire
multiple-test sequence,  is generated from
the data obtained from the test that is used
to determine whether the pipeline is  leak-
ing (in many instances these are the data
from the  last test in the sequence). This
histogram is needed to estimate the prob-
ability of detection. It may be a  simple
 matter to generate the histogram, or it may
 involve significant effort. The histogram of
 the signal-plus-noise may be measured
 directly  for each  leak rate of  interest by
 developing a histogram of the test results
 when a leak of a  given magnitude is
 present.  As an alternative, a  model that
 gives the relationship between the signal
 and the noise may be developed and vali-
 dated experimentally. If  the  relationship
 between the signal and  noise is known,
 the noise histogram  can  be used to esti-
 mate the signal-plus-noise histogram. This
 relationship can be difficult to develop un-
 less all sources of noise during the test are
 compensated for (or unless they are small).
 A model is required if one wants to know a
 system's performance at  many leak rates
 that are different from those specified in
 the EPA regulation.
   Step 11—Determine the system's
 sensitivity to trapped vapor.  The sensi-
 tivity of the leak detection system to vapor
 trapped in the  pipeline  system  must  be
 determined. To this end, three special leak
 detection tests will be performed. In each
 test, a specific amount of vapor, unknown
 to the tester, will be introduced into the line
 by means of an  apparatus especially de-
 signed for this purpose and described in
 the final report. These three tests will be
 intermixed with the other 25.
   Step 12—Conduct the performance
 analysis. The performance of the system
 in terms of PD at the EPA-specified leak
 rate and in terms of PFA is then  calculated.
 The protocol is designed so that the  PD
 and PFA of the system are determined with
 the manufacturer's threshold  at  the .leak
 rate and test pressure specified by the
 EPA regulation (i.e., 0.1, 0.2, or 3 gal/h). If
 the evaluation is not done at the pressure
 specified by the EPA, a method is given to
 calculate an equivalent leak rate at what-
 ever pressure is used. The protocol pro-
 vides a summary sheet to be  used in
 reporting a variety of other performance
 estimates so that the performance cari be
 compared to that of  other leak detection
 systems. If a system uses a multiple-test
 procedure, the protocol requires a second
 performance estimate based on noise and
signal-plus-noise data from the first test of
the multiple-test sequence.
   Step 13—Report the results. The fi-
 nal step is to report the results of the
 evaluation in  a set of standardized forms
found in an appendix to  the final report.
With the  information provided in  these
forms, the evaluation can be independently
reviewed and verified.

-------
 Reporting the Results
   One of the appendices of the complete
 final report contains a standard form on
 which the results of the evaluation can be
 summarized. The performance character-
 istics of the instrumentation, the estimates
 of the system's performance in detecting
 leaks in the ambient environment, and the
 sensitivity of the system to trapped vapor
 are summarized in a set of tables. The test
 conditions and pipeline systems to which
 the detector is applicable are  also  pre-
 sented. Seven  attachments to the form
 give additional details about the system
 and the evaluation. With the data and in-
 formation provided in these attachments,
 all of the results of the evaluation can be
 independently reviewed and verified. The
 seven attachments include:
 •   a description of the system
 •   a summary of its performance
 •   a summary of the configuration'of the
    pipeline system (s)
 •   a summary of product temperature
    conditions
 •   a summary of leak tests
 •   a summary of trapped vapor tests
 •   a summary of test  results  used to
    check the relationship supplied by the
    manufacturer for combining the signal
    and the noise.
   The full  report was submitted in fulfill-
 ment of Contract No. 68-03-3409 by Vista
 Research, Inc.,  under the sponsorship of
the U.S. Environmental Protection Agency.

-------

-------
  Joseph W. Maresca, Jr., Robert M. Smedfjeld, Richard F. Wise, and James W. Starr
   are with Vfefa Research, Inc., Mountain View, CA 94042.
  Anthony N. Tafurl is the EPA Project Officer (see below).
  The complete report, entitled "Standard Test Procedures for Evaluating Leak Detection
   Methods: Pipeline Leak Detection Systems," (Order No. PB91-106245/AS; Cost:
   $23.00, subject to change) will be available only from:
         National Technical Information Service
         5285 Port Royal Road
         Springfield, VA 22161
         Telephone: 703-487-4650
  The EPA Project Officer can be contacted at:
         Risk Reduction Engineering Laboratory
         U.S. Environmental Protection Agency
         Cincinnati, OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati, OH 45268
      BULK RATE
POSTAGE & FEES PAID
 EPA PERMIT NO. G-35
Official Business
Penalty for Private Use $300
EPA/600/S2-90/050

-------